A product photo exported at 4000 pixels wide and dropped straight into a WooCommerce media library gets served at that same size to every visitor, whether they're viewing it as a 300-pixel thumbnail on a category page or full-screen on a product sheet. The browser downloads the full file, decodes it, then scales it down in the layout: bandwidth spent for nothing. Multiply that by every image on every page view, and the gap between "images as uploaded" and "images as actually needed" becomes a recurring cost line, not a one-time inefficiency.
That cost shows up in two places. The first is direct: hosting and CDN bandwidth billed by the gigabyte, storage that keeps growing as a catalog expands, server CPU spent re-serving the same oversized file thousands of times. The second is indirect and usually larger: page load time. Largest Contentful Paint on a product page is very often the main product image itself, and a slow LCP measurably correlates with lower conversion rates and worse ranking signals in Google's Core Web Vitals. A few hundred extra milliseconds per page, multiplied across a catalog and a season of traffic, is not a rounding error for a store that depends on organic and paid traffic converting.
The fix isn't re-exporting every file by hand in the right size and format for every placement: that doesn't scale past a few dozen products, and it has to be redone every time a new device size or format gains adoption. The two mechanisms that actually solve this are structural rather than one-off: computing each needed variant (a given file, at a given size, in a given resize mode, in a given format) exactly once and serving it from a permanent disk cache afterward, and picking the best format automatically from what the visitor's browser actually announces it supports via its Accept header, rather than guessing or shipping the same JPEG to everyone.
This is precisely what ImagePipe does inside a WordPress/WooCommerce install: presets define the sizes a catalog actually needs, each variant is generated once and cached, and format negotiation serves AVIF or WebP to browsers that support them without touching the original files. It doesn't eliminate the cost of serving images (nothing does), but it removes the cost of serving the same image the same way, over and over, to browsers that could have received something smaller.